A study on applicability of fuzzy k-member clustering to privacy preserving pattern recognition
Hirohide Kasugai, Arina Kawano, Katsuhiro Honda, Akira Notsu · 2013
One of useful approaches in privacy preserving data mining is a priori data anonymization, in which each record are anonymized so that any records cannot be associated with a certain person. For effective data anonymization, clustering approaches have been applied. In a previous work, it was shown that a fuzzy clustering approach can achieve data anonymization without significant loss of information because it effectively merges similar records into clusters where each record is not distinguishable from others after within-cluster merging. This paper studies on the applicability of fuzzy k-member clustering to privacy preserving pattern recognition, in which the goal is to perform supervised pattern recognition keeping a certain anonymization level.